09 · Safety
Moderation
Before you show user content in your app, or send it to a model,
you can check it for harassment, hate, violence, self-harm and more.
omni-moderation-latest works on text and images, and
the endpoint is free to use.
Docs ↗
OpenAI does not let web pages call this endpoint directly, so the
request goes to moderation-proxy.php on this server,
which passes it on. That is how a real app calls the API, too.
No API key found. Go back to the overview to add one.
Result
Flagged means the model's score for a category was above OpenAI's threshold. Scores run from 0 to 1. The text and image tags show which input caused each score.